Intelligent Teaching Algorithm And Application Of Taekwondo Based On Deep Learning

Physical education is one of the important contents of implementing quality education in China,and it has a basic supporting role for improving the people’s physical fitness and the quality of life.In recent years,the rapid development of artificial intelligence technology has a significant impact on the content,mode,method and system of traditional physical education.How to apply artificial intelligence technology to physical education,break through the limitation of time and space scene and knowledge transfer mode in traditional mode,has become the key problem of Intelligent Physical Education in the era of artificial intelligence.Taking taekwondo as the research object,this paper explores the application of artificial intelligence technology in Taekwondo education.The reason why we choose Taekwondo is that taekwondo is popular because of its practicality and appreciation.At present,most colleges and universities in China have set up taekwondo courses,and Taekwondo has become an important part of college physical education.However,with the rapid increase of people practicing taekwondo,many problems are exposed in traditional education methods:(1)The number of students taking Taekwondo Courses is increasing,and the number of professional Taekwondo teachers is insufficient,which leads to the difficulty in ensuring the quality of students’ learning;(2)The traditional teaching method of Taekwondo is highly dependent on the coach,and the training of students alone cannot guarantee the training quality;(3)At present,the action evaluation relies on labor,and the results are easily affected by subjective factors such as the referee’s energy limitation,and cannot form an objective and accurate completion quality score.Therefore,it is necessary to introduce artificial intelligence technology to study a new Taekwondo teaching and training mode and method,to provide students with accurate,intelligent,and not subject to venue and time restrictions on movement guidance.In response to the above problems,this paper uses deep learning technology to develop a Taekwondo intelligent education system based on human pose estimation technology.The main goal of this system is to identify whether Taekwondo moves meet the standard,and give comprehensive evaluation and improvement suggestions on the quality of the completion of the moves,so as to get rid of the constraints of the coach and the venue,and realize automated and intelligent auxiliary teaching.Specific work content includes:(1)Taking Taekwondo’s poomsae as the research object,a set of intelligent evaluation algorithms for Taekwondo’s movement quality based on deep learning are proposed.Through multi-view feature extraction and modeling for the collected poomsae movement video,this method provides accurate and robust evaluation results.First,the multi-view video of athlete’s poomsae movement is collected by using multi-angel cameras,and the key joint point coordinates of human body is obtained by using a new human body gesture recognition model which fuses deep transfer learning technique.Second,according to the requirement of the Taekwondo training outline,a multi-angel feature extraction algorithm fusing Taekwondo professional knowledge is proposed.This algorithm numerically quantizes the action pose in each frame of the video to construct feature vector.Finally,the obtained samples are input into the long short-term memory neural network for training,while a new scoring index is proposed.The experiment is carried out on a set of actual collected Taekwondo videos.The results show that,compared with the traditional manual single-view evaluation mode,the proposer method can not only accurately identify the action gesture,but also effectively quantify the action intensity and objectively score the action quality.(2)This paper designs a new taekwondo intelligent education system based on video analysis.This system is supported by the Taekwondo action quality intelligent evaluation algorithm,which can automatically complete the action quality evaluation and provide training suggestions.The system is released on mobile devices.The front end of this system collects exercise video in real time and uploads it to backend server.By introducing deep learning techniques,the key points of human body posture can be recognized.Then referencing the demand of taekwondo training program,the key feature representation of corresponding actions can be extracted,followed by the intelligent classification and automatic discrimination for finishing quality of the actions.Finally,taking two typical actions “nailiecaki” and “hooking kick” as examples,this paper elaborates the implementation process of taekwondo action correction and intelligent scoring.